Add performance warning for MAP_FILTER

Author: abhinavmuk04

presto-main-base/src/main/java/com/facebook/presto/sql/analyzer/ExpressionAnalyzer.java

@@ -154,6 +154,7 @@
 import static com.facebook.presto.common.type.VarcharType.VARCHAR;
 import static com.facebook.presto.metadata.BuiltInTypeAndFunctionNamespaceManager.JAVA_BUILTIN_NAMESPACE;
 import static com.facebook.presto.spi.StandardErrorCode.OPERATOR_NOT_FOUND;
+import static com.facebook.presto.spi.StandardWarningCode.PERFORMANCE_WARNING;
 import static com.facebook.presto.spi.StandardWarningCode.SEMANTIC_WARNING;
 import static com.facebook.presto.sql.NodeUtils.getSortItemsFromOrderBy;
 import static com.facebook.presto.sql.analyzer.Analyzer.verifyNoAggregateWindowOrGroupingFunctions;
@@ -1110,6 +1111,15 @@ else if (frame.getType() == GROUPS) {
             FunctionHandle function = resolveFunction(sessionFunctions, transactionId, node, argumentTypes, functionAndTypeResolver);
             FunctionMetadata functionMetadata = functionAndTypeResolver.getFunctionMetadata(function);
 
+            String functionName = functionMetadata.getName().getObjectName().toUpperCase();
+            if (functionName.equals("MAP_FILTER")) {
+                String warningMessage = String.format(
+                        "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance.",
+                        functionName
+                );
+                warningCollector.add(new PrestoWarning(PERFORMANCE_WARNING, warningMessage));
+            }
+
             if (node.getOrderBy().isPresent()) {
                 for (SortItem sortItem : node.getOrderBy().get().getSortItems()) {
                     Type sortKeyType = process(sortItem.getSortKey(), context);

presto-main-base/src/test/java/com/facebook/presto/sql/analyzer/TestAnalyzer.java

@@ -152,6 +152,40 @@ void testNoORWarning()
         assertNoWarning(analyzeWithWarnings("SELECT * FROM t1 JOIN t2 ON t1.a = t2.a \n" + "AND (t1.b = t2.b OR t1.b > t2.b)"));
     }
 
+    @Test
+    public void testMapFilterWarnings()
+    {
+        assertHasWarning(
+                analyzeWithWarnings("SELECT map_filter(x, (k, v) -> v > 1) FROM (VALUES (map(ARRAY[1,2], ARRAY[2,3]))) AS t(x)"),
+                PERFORMANCE_WARNING,
+                "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance."
+        );
+
+        assertHasWarning(
+                analyzeWithWarnings("SELECT map_filter(x, (k, v) -> k = 2) FROM (VALUES (map(ARRAY[1,2,3], ARRAY[10,20,30]))) AS t(x)"),
+                PERFORMANCE_WARNING,
+                "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance."
+        );
+
+        assertHasWarning(
+                analyzeWithWarnings("SELECT map_filter(x, (k, v) -> k IN (1, 3)) FROM (VALUES (map(ARRAY[1,2,3], ARRAY[10,20,30]))) AS t(x)"),
+                PERFORMANCE_WARNING,
+                "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance."
+        );
+
+        assertHasWarning(
+                analyzeWithWarnings("SELECT map_filter(x, (k, v) -> v IN (20, 30)) FROM (VALUES (map(ARRAY[1,2,3], ARRAY[10,20,30]))) AS t(x)"),
+                PERFORMANCE_WARNING,
+                "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance."
+        );
+
+        assertHasWarning(
+                analyzeWithWarnings("SELECT map_filter(x, (k, v) -> k + v > 25) FROM (VALUES (map(ARRAY[1,2,3], ARRAY[10,20,30]))) AS t(x)"),
+                PERFORMANCE_WARNING,
+                "Using MAP_FILTER is expensive and may slow down training pipelines. Consider using map_subset or other built-ins for better performance."
+        );
+    }
+
     @Test
     public void testIgnoreNullWarning()
     {